The Use of Twin Families in Quantitative Trait Locus Mapping

Wei J. Chen ； Pao-Hsiu Kuo

《Journal of Genetics and Molecular Biology》 11卷2期 (2000/06) Pp. 89-91

https://doi.org/10.30047/JGMB.200006.0014

並列摘要

Twin studies have been instrumental in elucidating gross genetic and environmental factors on phenotypes that exhibit quantitative or continuous variation in the population at large. However, twin studies have not been used extensively in mapping specific or individual genes that influence quantitative traits. The importance of quantitative trait locus (QTL) approach to complex disease lies in the fact that risk factors of such a complex disease may also be genetically influenced. For many dichotomous traits, there exits one or more continuously distributed liability scales or strong risk factors that may lend more power to detect relevant QTLs than the complex trait itself. This presentation will address the use of twin families in QTL mapping. To begin with, 1 will review the recent developments in methodology for mapping QTL. The traditional approach to linkage detection of quantitative traits was based on a regressional method outlined by Haseman and Elston (1972). Despite its elegance and simplicity, the Haseman-Elston procedure has a number of shortcomings and has been extended in several aspects. First, multipoint mapping procedures were developed to incorporate the increasingly available DNA markers. Second, by selecting the sib pairs who are extremely concordant or extremely discordant for the quantitative phenotype greatly enhances the power of detection of QTLs in humans. Third, the analysis of multivariate rather than univariate sib-pair data has been shown to increase the power to detect a QTL substantially. Finally, a genetic covariance-structure modeling instead of examining the phenotype difference only has been developed to detect QTL in sib pair data. Compared to the traditional Haseman-Elston regression approach, genetic covariance structure modeling approach has been shown to be more powerful and more flexible in incorporating multivariate date and general pedigree. Following this line of thinking, twins are ideal samples for mapping QTLs. Since DZ twins can be considered age-matched sibling pairs and MZ twins can be considered as sharing all alleles identical by descent (i.e.,=1), their use in sib-pair-based gene mapping analyses mentioned above is ideal. Furthermore, since one can exploit the contrast between MZ and DZ twins to better characterize the effect of genetic and environmental factors on variation in a trait, one could conceivably exploit this contrast to better characterize the role of a particular locus with respect to variation in a phenotype. If a multivariate approach is adopted, the power to detect QTL is enhanced further. Such an multivariate QTL linkage analysis using twin data now can be easily carried by a package program Mx (Neale, 1997). It is worthwhile to note that in order to have a more accurate estimate of in DZ twins, it is better to include twins’ parents in the genotyping. One further advantaged in recruiting twins’ parents is that the parents phenotype can be easily incorporate in the covariance structural modeling to allow for parent-offspring correlation in the model. That is, not only twins but also their parents can be incorporated in the analysis to fully exploit the information contained in a twin-family. There have only few studies that have examined the power of the covariance structural modeling approach to twin-based gene mapping. Martin et al. (1997) and Boomsma and Dolan (2000) have investigated the sample size required to detect a QTL effect using a likelihood ratio test. In their approach, a model consisting of a QTL effect (σq2), residual additive genetic effect (σa2), and random environment effect (σe2) was compared to a reduced model with the exclusion f σq2. However, the QTL effect and additive genetic effect are completely confounded in MZ twins and thus do not provide any information in this approach of power consideration. Thus, MZ twins might be treated as DZ twins with =1 in empirical data analysis. Following this approach, we have employed Mx program to compute the sample size required for the detection of a QTL effect under various conditions. The assumptions used in our calculations are :1) the QTL has 4 equal-frequent alleles; and 2) the significance level was set at 0.05. The results are summarized in the table. As can be seen from the table, a sample size of 400 pairs has a reasonable power to detect a QTL that is responsible for 30% or more of trait variance. If the variance due to QTL is less than 30%, then it is difficult to detect its existence given the size of 400 pairs. Nevertheless, the power will increase further if we employ two or more correlated measured for a phenotype and a multivariate approach is adopted in the analysis (Martin et al., 1997). Alternatively, Schork and Xu (2000) have assessed the power of twin studies in QTL mapping by employing a profile likelihood ratio test. Although the scenarios examined were limited and many remains to be investigated, their results do indicate that incorporating MZ twins in the profile likelihood ratio test increase power considerably. Empirically, Zhu et al. (1999) studied 153 pairs of MZ twins and 199 pairs of DZ twins and successfully detected a ZTL for mole density that explains around 30% of the trait variation. Therefore, from both power calculations and previous empirical studies, twins have a reasonable power for the detection of QTL pertaining to many complex phenotypes that are of importance in medicine and public health.